Farmers’ Perceptions of Agricultural Land Values in Rural
Pakistan
Shehryar Rashid
Asjad Tariq Sheikh
About the Author(s)
Shehryar Rashid is a Research Analyst at the Pakistan Strategy Support Program (PSSP) of the
International Food Policy Research Institute (IFPRI)
Asjad Tariq Sheikh is a Senior Research Assistant at Innovative Development Strategies (IDS)
Note: Special thanks to Dr. Stephen Prescott Davies, Dr. Hina Nazli, Dr. Sarfaraz Khan Qureshi, Dr.
David Spielman, Dr. John Mellor, Dr. Katrina Kosec, Mr. Syed Hamza Haider, Ms. Faryal Ahmed, and Ms.
Hira Channa for their comments and editing.
Farmers’ Perceptions of Agricultural Land Values in Rural
Pakistan
Abstract:
Land is an important social and financial asset in Pakistan. Identifying the determinants
of land value is important because these determinants have an impact on the utility of the land. In
rural areas, utility can be defined in terms of agriculture productivity. Therefore, the Government
of Pakistan will benefit from research which determines what characteristics affect the value of
land which will also affect agriculture productivity. This research study will use a hedonic
regression model and two-stage least square method to determine what demographic, site,
development, or physical characteristics have a correlation with the perceived value per acre of
rural agricultural land in Pakistan. Data for this study was obtained from the Pakistan Strategy
Support Program’s Rural Household Panel Survey (RHPS) conducted in 2012. Results from the
survey indicate that there is a high level of inequality in land ownership in Pakistan which can
limit opportunities for growth. Regression results indicate that most of the site and physical
characteristics are correlated with perceived land value and only a few of the development
indicators and none of the demographic variables have a statistical impact. Specifically fertile
land, lack of soil erosion, number of canal and ground water irrigations, location of plot at head
and middle of watercourse, access to electricity, internal road, cotton grower, sugarcane grower,
and average mauza income are positively correlated with perceived land value per acre.
Waterlogging, salinity, and distance to nearest city have a negative correlation. These results
suggest that the Government of Pakistan should support regional initiatives which improve
access to good quality water, improve the condition of land used for agriculture, and provide
access to electricity.
Table of Contents
I.
Introduction .......................................................................................................................................... 4
II.
Literature Review .................................................................................................................................. 5
III.
Methodology..................................................................................................................................... 8
IV.
Sample Characteristics ...................................................................................................................... 9
V.
Model .................................................................................................................................................. 16
VI.
Results ............................................................................................................................................. 18
VII.
Conclusion and Policy Implications ................................................................................................. 22
VIII.
References ...................................................................................................................................... 24
IX.
Appendix ......................................................................................................................................... 26
List of Tables
Table 1: Gini Coefficient for Land Ownership in Pakistan from 1972 to 2012 ........................................... 15
Table 2: Summary Statistics for Variables used in the Model (Dependent variable: Land value (Rs/acre))
.................................................................................................................................................................... 17
Table 4: Description of Sample from Round 1.5 of PSSP Rural Household Survey..................................... 26
Table 5: Distribution of Residuals ............................................................................................................... 31
List of Figures
Figure 1: Perceived Value of Land (Rupees / Acre) ....................................................................................... 9
Figure 2: Percentage of Farmers Experiencing Waterlogging by Province ................................................. 10
Figure 3: Percentage of Farmers Experiencing Salinity by Province ........................................................... 10
Figure 4: Perceived Value of Land per District (Rupees per Acre) .............................................................. 11
Figure 5: Tenancy Status by Plot ................................................................................................................. 12
Figure 6: Tenancy Status by Province ......................................................................................................... 12
Figure 7: Farmer Categories ........................................................................................................................ 13
Figure 8: Sub Categories of Marginal Farmers............................................................................................ 14
Figure 9: Average Land Holding by Farmer Category ................................................................................. 14
Figure 10: Tenancy Status for Farmer Categories....................................................................................... 15
Figure 11: Gini Coefficient for Pakistan ...................................................................................................... 16
I.
Introduction
Pakistan’s agriculture sector is crucial because it is responsible for providing food,
shelter, and clothing to a massive population of 180 million people which is growing at a rate of
2 percent per annum. Land is a valuable asset and a symbol of prestige for the rural population in
Pakistan. According to the recent Pakistan Economic Survey of 2012-13, the agriculture sector
contributes around 21% to GDP and provides employment for around 45% of the work force,
who are primarily based in rural areas. According to the Agricultural Statistics of Pakistan of
2010-11, the total geographic area of Pakistan is approximately 79.6 million hectares. Around
27.7 percent of Pakistan’s land is currently under cultivation and the cultivatable waste lands
offer good possibilities for crop production. The total cropped area of Pakistan increased from
21.82 million hectares in 1990-91 to 22.72 million hectares in 2010-11 (Agricultural Statistics of
Pakistan 2010-11) and the total population of Pakistan increased from 118 million to 175 million
during the same time period. Similarly the tenancy status of land management and land
ownership pattern has changed over time. For example, large landowners are shifting their
preferences from managing their land on their own towards leasing or sharecropping the land to
be managed by others (Agricultural Census 2010).
Land is a difficult resource to exchange because of certain constraints such as the fact
that land is immobile and there may be significant differences in the quality of land.
Additionally, appropriate institutions may not exist which allow for costless exchange of land.
Land is a finite resource and ideally the market with demand and supply forces should be able to
determine the equilibrium price. However, this is not the case in Pakistan where land markets
mostly don’t exist at a formal level and the value of land is being priced arbitrarily and without
any scientific backing. In some cases the price of land is being influenced by large landowners.
Furthermore, in Pakistan, there is no appropriate or historical collection of data on land buying /
selling and land revenue (provincial revenue departments are supposed to maintain records of
land ownership, however, this data is usually not publicly available). For example, there is no
nationally representative survey on household land purchases and only the recently released
Pakistan Agriculture Census of 2010 included some data on change in land ownership patterns.
Lack of a formal land market and sufficient data means it is difficult to identify the
determinants of value for land in Pakistan. Determinants of land value have an impact on the
utility of land. In rural areas, the utility of land can be determined as the productivity level of
agriculture land.
In Pakistan, rural land in the agriculture sector is important because most of Pakistan’s
land can be classified as rural and is based in the agriculture sector. Since Pakistan is a
developing country, the Government is trying to implement policies which promote development
and reduce poverty. This can be done by promoting investment and policies which increase
agriculture productivity. Yet the same polices and investments which increase agriculture
productivity indirectly also increase agricultural land prices (Gardner et al 1979). Previous
literature has shown that land ownership and the productivity level of agricultural land are very
closely related to poverty and development (Deininger 2004 and Hirshima 2008). Finding out
what factors affect land values in rural Pakistan could help the Government of Pakistan decide
what to invest in to promote development of rural land. Similarly, proper investment into rural
areas can turn them into centers of commerce which will boost productivity and economic
growth. In the long run, this will improve competition in the area.
According to the Food and Agriculture Organization (2003), price of land is one of the
tools which can be used to manage land resources. Price of land itself is important because it
reflects the level of government reforms which are used to support agricultural production.
However, studying land itself is difficult because land value has different definitions and land
markets in Pakistan do not exist at a formal level. In order to resolve the problem of the
definition of land value, we will be using the perceived value of the land by the farmers who
manage the land.
Previously, many studies revealed that there is a positive impact of attributes/
characteristics of land on the value of agricultural land (Vasquez et al. (2002), Guiling et al.
(2009), Cavailhès and Wavresky (2003), Peterson (1984 and 1986)). However, no such study
exists for Pakistan. Specifically, this research study will look at the relationship between physical
and economic characteristics and whether they are correlated with property values in rural
Pakistan. Due to a lack of suitable and reliable data, we used perceived value of land per acre as
our dependent variable. Specifically, we asked the farmers managing the land what is their
perceived value of the land they are managing. The rest of the paper is organized in the following
manner; section II gives a literature review on the subject, section III describes the methodology
used and Section IV gives data on the sample. Section V describes the model we will be using to
examine the relationship and Section VI provides results. Specifically we will be using a hedonic
regression model based on the approach originally presented by Bover and Velilla (2002).
Section VII is a conclusion along with brief policy recommendations.
II.
Literature Review
As mentioned above, land is an important social and financial asset, yet there is a high
level of inequality of land ownership in Pakistan. For example, the Household Income and
Expenditure Survey of 2001-2002 stated that 43.13% of households in Pakistan were in rural
areas. Out of the rural households 24.02% were landless, 42.27% owned less than 5 acres,
22.40% owned 5 to fewer than 12.5 acres, and 11.31% owned 12.5 acres or above. According to
Qureshi and Qureshi (2004), the Gini coefficient for land ownership in Pakistan significantly
increased from 0.66 in 1972 to 0.75 in 2000. Highest increase in inequality of land ownership is
seen in the province of Punjab from 0.63 to 0.71 and KPK (NWFP at the time) from 0.68 to 0.86.
Gini Coefficient is almost the same for Sindh 0.69 to 0.67 and Balochistan 0.69 to 0.68.
Similarly, Mumtaz and Noshirwani (2006) performed a mapping exercise in 3 provinces (Punjab,
Sindh, and KPK) and found that 40% of rural land is owned by 2.5% of households. They also
found that women prioritized inheritance as an issue that bothered them. Women faced issues
that they were manipulated out of their inheritance, had to forfeit their share in favor of brother
or son, and were unable to pursue inheritance in court.
The Government of Pakistan has tried on three different occasions (1959, 1972, and
1979) to implement land reforms to solve problems with land usage and land development in
Pakistan. PANOS (2011) stated that previous attempts at land redistribution have failed because
of fragmentation which is hurting agriculture output. Ownership of land is rarely registered
(despite law making land ownership registration mandatory) and is passed on through
inheritance. An estimated 40% of cases brought before lower level civil courts and high courts
are land related disputes (Aftab et. al. 2012). On August 10, 1989 the Supreme Court Shariat
Appellate Bench declared that a maximum ceiling for land holding was illegal as per Islamic
Law. Therefore, in recent decades, the focus had shifted from land redistribution towards
improving records of land ownership.
Hirashima (2008) showed that the price of land in the province of Punjab in Pakistan and
India was increasing at a faster rate than rent. The basic reason for this he argued is that the
demand for land in Pakistan is price inelastic because of its importance to social status and the
inheritance law. He argues that even though land is a factor of production just like labor and
capital, land is significantly different because it is not man made and has limited scope of
extension.
Renkow (1993) showed that in both irrigated and rain fed areas of Punjab, productivity
increases have led to greater returns to land in the form of higher real land rents and higher real
land prices. Evidence was collected from the wheat producing province of Punjab. Other factors
positively affecting land prices are remittances from abroad, mechanization, improved usage of
wheat seed varieties, fertilizer usage, and changes in agronomic practices of rain fed areas.
It is worth mentioning here that apart from these studies mentioned above, not much
work has been done to examine what affects land prices specifically in Pakistan. Most of the
work on land in Pakistan such as Naqvi et al. (1989 and 1987) has focused more on the impact of
land reforms. The Federal Bureau of Statistics does publish agricultural statistics and agricultural
census data which described changes in land usage and ownership patterns. The most recent
Pakistan Agricultural Census was completed in 2010.
It is far more common to find international literature examining factors affecting land
prices. In this case, a far more widely used approach is done using hedonic modeling. The basis
for a hedonic pricing model can be found in Rosen (1974) and this model can be used to estimate
the impact of a range of characteristics such and economic, environmental, and location variables
and how they affect the price of goods. In this case, the assumption is that consumers value the
characteristic of goods or the services they produce rather than the goods themselves. However,
no such study has been found examining land values in Pakistan.
For example Peterson (1984 and 1986) used a hedonic regression model to analyze land
prices in Africa and Europe. The author found that 70% of the variation in land prices was due to
non-farm factors such as precipitation. Taylor and Brester (2005) look at the impact of a noncash
income transfer program on agricultural land values. Specifically they use a hedonic regression
model to look at the impact of a sugar program on agricultural land values in Montana. They find
that noncash income transfers have a positive impact on land values. Similarly Roberts et. al
(2003) provide examples where government cash transfers or other government programs can
have a positive impact on land prices. Bover and Velilla (2002) also use a hedonic price model to
determine if quality indicators such as location and floor size affect land values of multi-unit
housing in various cities in Spain. Results by city vary, however the hedonic regression results
indicate that overall there is a positive relationship. Vural and Fidan (2009) provide further
evidence while using a hedonic price model studying the effects of factors affecting land prices
in Turkey. Results indicate a high correlation between type of organic matter in the area and land
size. Saita (2003) uses a similar approach in examining factors affecting land values during
auctions in Tokyo. The author’s results contradict results mentioned earlier mainly because the
housing market bubble in Tokyo had collapsed around the time. The author finds that land prices
respond mostly to market conditions.
Researchers that did not use a hedonic model approach to study land values include
Vasquez et al. (2002) who used data on 453 sales from the years 1993-1994 for Idaho farmland
prices. They found that development variables had an effect on land prices compared to region
specific data. Similarly Cavailhès and Wavresky (2003) also analyzed the impact of urban
influence on farmland prices. They found that farmland prices fall sharply close to the city and
then fell slowly the further you move away. Their data consisted of more than 2,000 sales in
Dijon and the surrounding region in France. Guiling et al. (2009) examine the impact of urban
proximity on agricultural land values using Oklahoma’s agricultural land data. In this case urban
proximity is defined in terms of population, time and real income. The authors determine that
population and income have more of an impact on agriculture land values whereas distance does
not seem to have an impact meaning that there is no preference towards living closer to a city.
Plantinga and Miller (2001) find evidence from New York State that future land
development has a positive effect on current land values. The implication here is that perceived
future development will cause perceived rent on agriculture land to be higher and therefore the
price of current agriculture land will increase.
Gardner and Nuckton (1979) describe how land prices increased rapidly in the United
States from the 1950s to the 1970s. The authors attribute this increase in land price to an increase
in income from land ownership, increase in productivity of agriculture, government support, and
other income support programs. Other factors affecting land values is rapid urbanization and
foreign investment from companies entering the market.
Deininger et. al. (2014) also acknowledged that while the importance of land governance
in agricultural growth and development has been acknowledged, the extent to which it has been
effectively administered or implemented over time has not been successful in African countries.
Deininger argues in favor of using the Land Governance Assessment Framework (LGAF) as a
diagnostic tool (developed by the World Bank) used to identify how land governance affects
productivity. Using such an approach allows for transferring policy recommendations into a
more valued impact on agricultural productivity.
Ciaian et. al. (2012) attempt to examine institutional factors affecting rent and land
values of agricultural land. The authors find that agriculture policy promoting income, land
markets, institutions, and regulations, type of land usage and social capital (government fixing
land prices) has a positive impact. Other factors such as high transaction costs and credit market
constraints have a negative impact on land prices. Similarly, Du and Mulley (2007) find that the
introduction of a public rail system can have a positive impact on land values however the
change occurs over a longer period of time.
Stillman (2005) examines changes in the value of land in New Zealand from 1989 to
2003. Specifically the authors examines if profitability of land, alternative land usages, local
climate, and local conditions has an impact on land values. The author finds that value of rural
land increased a great deal during the mentioned period. This increase can largely be attributed
towards climate, local conditions, and initial usage of land.
III.
Methodology
The Pakistan Strategy Support Program (PSSP) recently completed two related rounds
(known as round 1.0 and round 1.5) of a rural household survey in 2012 in which 2,090
households from 19 districts across Pakistan were interviewed. These 19 districts included 12
from Punjab, 5 from Sindh and 2 from Khyber-Pakhtun-Khwa (KPK)1. Round 1.0 was a multitopic survey which included questions from different economic areas and Round 1.5 was a
survey specifically focused on agriculture. Therefore the sample of Round 1.5 only included
households from Round 1.0 who were involved in farming (942 households). For a detailed
description of the sample please refer to Table 4 in the Appendix section. This survey is known
as the Rural Household Panel Survey (RHPS).
This paper will utilize the data from the PSSP’s rural household survey. Specifically the
paper will use data relevant for land valuation from Round 1.0 by using community level data
1
Note: Balochistan was removed from the sample due to security reasons.
and data from Round 1.5 of the survey. Following the literature review, this study will try to fill
in a gap in the current literature by examining what factors affect land prices in rural Pakistan. A
selection of variables (physical and economic), which theoretically have an impact on land
values, will act as independent variables. Specifically we will be using a hedonic regression
model and two-stage least square model approach which has not been used previously for
studying land prices in Pakistan. Section IV below provides sample characteristics and Section V
will describe the model in more detail.
IV.
Sample Characteristics
The sample we will be using has 942 households out of which 521 are in Punjab, 305 are
in Sindh, and 116 are in KPK. Figure 1 below provides data on perceived value of land per acre
in this sample. In this case the household was asked about the perceived value of the agricultural
land (Rs./Acre) if it was sold today.
Average Land Value (Rs./Acre)
Figure 1: Perceived Value of Land (Rupees / Acre)
900000
800000
700000
600000
500000
400000
300000
200000
100000
0
Punjab
Sindh
KPK
Provinces
The data indicates that the self-reported value of land per acre is highest in KPK at
892,115 Rs/acre. This is followed by Punjab with a perceived value of land at 874,439 Rs/acre.
In Sindh the perceived value of land per acre is much lower at 319,650 Rs/acre. Lower perceived
value in Sindh can largely be explained by physical characteristics such as a larger proportion of
salinity and water logging issues. These households will have a lower perceived value of the land
compared with other households located in Punjab or KPK. Another possible reason for lower
perceived value of land is that the management of labor is not as efficient in Sindh compared to
Punjab and KPK.
Figure 2: Percentage of Farmers Experiencing Waterlogging by Province
The figure above provides data on the percentage of farmers experiencing water logging
by province. As expected, the province of Sindh suffered the most with 42.6% of the farmers
stating that they experienced waterlogging on their plot(s). In Punjab 5.6% of the farmers and in
KPK 7.8% of the farmers noted that they experienced water logging on their plot(s).
Figure 3: Percentage of Farmers Experiencing Salinity by Province
The figure above provides data on the percentage of farmers experiencing salinity by
province. Once again, the province of Sindh fared the worst with 37% of the farmers responding
that they experienced problems with salinity on their plot(s). Only 4.0% of the farmers in Punjab
and 8.6% of the farmers in KPK responded that they experienced problems with salinity. Figures
for waterlogging and salinity can be used to explain lower perceived value of land in Sindh. It
must be noted that data was collected in 2012 only a year after the devastating floods which
affected agricultural land in Sindh.
Figure 4 below disaggregates the data further into districts and provides the perceived
value of land per acre by district.
Figure 4: Perceived Value of Land per District (Rupees per Acre)
Average Land Value (Rs./Acre)
1400000
1200000
1000000
800000
600000
400000
200000
0
Districts
The results indicate that in terms of perception, the most expensive land is in Rahim Yar
Khan at 1,306,863 Rs/acre. The least expensive agricultural land is in the district of Hyderabad2
at 135,583 Rs/acre.
Figure 5 below provides the tenancy status of each of the plots in the sample. The 942
households in the sample owned a total of 1,296 plots.
2
Note that the sample for the PSSP Rural Household Panel Survey was creating using data from the most recently
available Census of 1998. Since then the district of Hyderabad has been divided into 4 districts known as
Hyderabad, Tando Muhammad Khan, Tando Allahyar, and Mitiari. Villages in the sample are located outside
current day Hyderabad district.
Figure 5: Tenancy Status by Plot
Percentage of Plots
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
Own
Rent In
Sharecrop in
Mortgage but
Managed by self
Tenancy Status of Plots
Results indicate that most of the plots (797 plots or 61.5%) are managed and owned by
the households. In 168 cases (13%) the plot is rented by the household for a fixed rent per month
or year. In 328 cases (25.3%) the plot is being managed by the household on a sharecrop basis
with a specific percentage going to the owner. In 3 cases (.2%) the plot has been mortgaged by
the household but is currently being managed by the household.
Figure 6 below disaggregates the data further and provides the tenancy status by province
for each plot in the sample.
Figure 6: Tenancy Status by Province
Percentage of Farmers
80.0
70.0
60.0
50.0
40.0
Pakistan
30.0
Punjab
20.0
Sindh
10.0
Khyber Pakhtunkhwa
0.0
Own
Rent In
Sharecrop in
Tenancy Status by Province
Mortgaged
but Managed
by self
The figure above shows that the tenancy status of the plots follows a similar distribution
for the province of Punjab and KPK. In Punjab 73.7% of the plots are owned by the households
themselves and this is the case for 66.8% of the plots in KPK. However the situation is different
in the province of Sindh. In Sindh only 33.5% of the plots are owned by the households and most
of the plots (65.4%) are being operated on a sharecropping basis by the household. The number
of plots being operated on a sharecropping basis is only 8.4% in Punjab and 15.9% in KPK.
Figure 7 below gives a description on the type of farmers managing the plots in the
sample. A marginal farmer is defined as a household which manages land less than 5 acres. A
small farmer manages greater than or equal to 5 acres but less than 12.5 acres. A medium farmer
manages greater than or equal to 12.5 acres but less than 25 acres. A large farmer manages 25
acres or above. Results from the RHPS sample are compared with the results from the Pakistan
Agricultural Census of 2010.
Figure 7: Farmer Categories
Percentage of Farmers
70.00
60.00
50.00
40.00
30.00
20.00
10.00
0.00
Marginal Farmers
Small Farmers
Medium Farmers
Large Farmers
Farmers Categories by Operated Land
PSSP/IFPRI
Agriculture Census-2010
Results indicate that in most cases (63.8%) the households can be classified as marginal
farmers. Another 27.6% households can be described as small farmers. A further 6.8 % of
households can be described as medium farmers and the remaining 1.8 % households can be
described as large farmers. Additionally, results by farmer category are comparable across the
RHPS sample and the Pakistan Agricultural Census of 2010. The figure below provides more
details for the farmer categories. This figure provides the average number of acres owned by
each farmer category group.
Figure 8: Sub Categories of Marginal Farmers
Percentage of Farmers
40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
<1
>= 1 - < 2
>= 2 - < 5
Farmers Categories by Operated Land
Figure 8 above divides the marginal farmers into sub categories. Here we are trying to see
some facts in the distribution of the amount of acres owned by marginal farmers. The three
categories are ownership of less than 1 acre, greater than or equal to 1 acre and less than 2 acres,
and greater than or equal to 2 acres and less than 5 acres. The data shows that most marginal
farmers (39.6% of the total number of farmers in the sample) own greater than or equal to 2 acres
and less than 5 acres.
Figure 9: Average Land Holding by Farmer Category
Average Land Holding (Acres)
45
40
35
30
25
20
15
10
5
0
Marginal Farmers
Small Farmers
Medium Farmers
Farmers Categories by Operated Land
PSSP/IFPRI
Agriculture Census-2010
Large Farmers
Results for the figure above show that on average, marginal farmers own 2.22 acres of
land, small farmers own 7.37 acres of land, medium farmers own 16.48 acres of land, and large
farmers own 38.71 acres of land. Figure 7 and figure 9 provide proof that there is a large level of
inequality for land ownership in Pakistan. Results also show that data from IFPRI’s RHPS
survey are comparable with the Pakistan Agricultural Census of 2010.
Average Farm Size (Acres)
Figure 10: Tenancy Status for Farmer Categories
45
40
35
30
25
20
15
10
5
0
Marginal Farmers
Small Farmers
Medium Farmers
Large Farmers
Own
Rent In
Sharecrop in
Mortgage but
Managed by
self
Tenancy status by Farmers Categories
Figure 10 above shows the disaggregated data on average farm size by tenancy status of
the plots and compare among farmers categories. For example, the figure above shows that the
average farm size for large farmers who own land is 36 acres. Similarly the average farm size for
medium farmers who own land is 16.22 acres, 7.16 acres for small farmers, and 1.73 acres for
marginal farmers.
Data from the figure above once again proves that the distribution of land ownership is
highly unequal with a small amount of households owning a large proportion of the rural
agricultural land. Using data from the sample, we were able to calculate Gini coefficients for
land ownership by households and compare results with earlier findings from Qureshi et al.
(2004). Note that ownership is defined in terms of plots which are in the household’s name
which means that plots that were rented out or are being operated on sharecropping basis were
attributed to the original owner.
Table 1: Gini Coefficient for Land Ownership in Pakistan from 1972 to 2012
Province / National
Pakistan
Punjab
Qureshi et al. 2004
1972
0.66
0.63
1980
0.65
0.62
PSSP 2012*
1990
0.66
0.62
2000
0.75
0.71
2012
0.68
0.61
KPK
Sindh
Balochistan
0.68
0.69
0.69
0.69
0.63
0.68
0.65
0.63
0.7
0.86
0.67
0.68
0.60
0.76
NA**
*Authors own calculation **NA = Did not survey due to security reasons
Qureshi et al. 2004 showed that the Gini coefficient appeared to be rising in Pakistan
overall from 1972 till the year 2000. A rising Gini coefficient implied that the inequality of land
ownership appeared to be increasing during this time. Our calculation for the Gini coefficient is
lower for Pakistan overall (0.59) and for each province. However this does not necessarily
indicate that land ownership inequality is decreasing because both studies used separate data
sources to calculate the Gini coefficient. Qureshi et al. (2004) used data from the Agricultural
Census Reports which had a larger sample size and covered a larger number of districts. We used
data from the PSSP’s Rural Household Panel Survey. The PSSP’s Rural Household Panel Survey
excluded a few districts from KPK for security reasons. Additionally our sample did not cover
the province of Balochistan.
Figure 11 below gives a graphic representation of how the Gini Coefficient was
calculated from our sample. The red line is the line of equality (each household owns the same
amount of land) and the blue line is the actual land ownership pattern. The Gini coefficient is
calculated as the area of “A” divided by the area of “A” plus “B” (Gini = A / (A+B)).
Figure 11: Gini Coefficient for Pakistan
3500
Number of Acres
3000
2500
2000
Ownership
1500
A
1000
Equality
500
B
1
38
75
112
149
186
223
260
297
334
371
408
445
482
519
556
593
630
667
704
741
778
815
852
889
926
0
Number of Households
V.
Model
We will be using a hedonic regression model and two-stage least square model to analyze
the mentioned relationship. The advantage of a hedonic regression model is that it divides the
explanatory variables into constituent parts and allows for analysis of different attributes
(example physical variables vs economic variables) on the dependent variable. We will be using
cross sectional data for the Pakistan’s Strategy Support Program’s Rural Household Survey from
the year 2012. Based on the approach by Bover and Velilla (2002), we used a model with a
theoretical form provided below:
Where is the log of perceived value of land per acre and
to
are a set of dummy
variables to calculate the effect of specific demographic or physical characteristics and
to
are the log of specific demographic, location, or development variables. This model can be
considered as a log-log model for continuous variables and not for other types of variables (ex.
dummy variables). The independent variables in the model can be categorized into four
categories which are demographic variables, site characteristics, development variables, and
location variables. The difference between site characteristics and location variables is that
location variables are usually fixed for the entire village and surrounding area and cannot be
changed. Site characteristics can differ between each plot. Development variables capture the
socio-economic wellbeing of the residents of the mouza. There are four different versions of the
model and the first two are standard hedonic regression models where one considers the impact
of renting a plot and the other considers the actual value of rent per acre. The next two models
used a two-staged least squares approach in order to counter potential issues with endogeneity in
which we used proxies to capture the effect of a change in wealth or development in a village.
Theoretically it is safe to assume that villages with a higher level of income are more likely to
travel longer distances, and therefore these variables can be used as proxies for average
agricultural income. Variables such as distance to nearest bank, city, and market are meant to
capture the effect of a change in the level of income of a village.
Summary statistics for the variables used in the model are provided in the table below.
Table 2: Summary Statistics for Variables used in the Model (Dependent variable: Land value (Rs/acre))
Variables
Land Value (Rs. / Acre)
Age of Respondent
Value of Rent
Average Mauza Income
Ever Attended School
Dummy for Ownership of Plot
Dummy for Renting in Plot
Dummy for Flat Land
Obs.
1296
1296
1296
1296
1296
1296
1296
1296
Mean
724023.9
41.80324
2688
168779.6
0.5933642
0.617284
0.1296296
0.7091049
Std. Dev.
634639
13.71747
8317.84
146951.9
0.4913954
0.486238
0.336025
0.4543505
Min
16000
14
0.01
0.01
0
0
0
0
Max
5200000
92
75000
917090
1
1
1
1
Variables
Dummy for Fertile Land
Dummy for Moderate Fertile Land
Dummy for No soil erosion
Dummy for Mild Soil Erosion
Dummy for Salinity
Dummy for Waterlogging
Number of Canal Irrigations
Number of Ground Water Irrigations
Dummy for Plot at Head
Dummy for Plot at Middle
Dummy for Village Electrification
Dummy for Internal Road
Dummy for Cotton Grower
Dummy for Rice Grower
Dummy for Sugarcane Grower
Distance to Nearest City
Distance to Nearest Tehsil Katcheri
Distance to Nearest Bank
Distance to Nearest District Katchari
VI.
Obs.
1296
1296
1296
1296
1296
1296
1296
1296
1296
1296
1296
1296
1296
1296
1296
1296
1213
1296
1296
Mean
0.1589506
0.7908951
0.8333333
0.1466049
0.121142
0.1535494
10.02627
8.414097
0.087963
0.2214506
0.9128086
0.2932099
0.0864968
0.1220263
0.0306345
12.7284
3.04642
13.0463
42.77932
Std. Dev.
0.3657712
0.4068265
0.3728219
0.3538482
0.3264182
0.3606554
11.88117
10.91017
0.2833504
0.4153834
0.2822242
0.4554096
0.142917
0.194041
0.108021
7.644005
0.5749389
8.247837
23.49335
Min
0
0
0
0
0
0
0.01
0.01
0
0
0
0
0
0
0
1
0.7
0
10
Max
1
1
1
1
1
0
77
60
1
1
1
1
1
1
1
35
4.32
35
115
Results
Table 3 below provides the results from the hedonic regression models described above.
Variables
Constant
Rent
Average Mauza Income
Ownership of Plot
Age of Respondent
Ever Attended School
Flat Land
Fertile Land
Moderately Fertile Land
No Soil Erosion
Mild Soil Erosion
Waterlogging
Salinity
Number of Canal
Log-Log (Dummy for
Rent in)
12.70***
(0.502)
-0.0262
(0.0789)
0.0135
(0.0129)
0.00320
(0.0619)
-0.0541
(0.0616)
-0.0174
(0.0460)
-0.000108
(0.0529)
0.323***
(0.114)
0.133
(0.0985)
0.589***
(0.147)
0.480***
(0.148)
-0.152*
(0.0913)
-0.125
(0.0930)
0.0397***
Log-Log (Value of
Rent-in)
12.69***
(0.499)
-0.00104
(0.00549)
0.0136
(0.0129)
0.00921
(0.0620)
-0.0541
(0.0616)
-0.0178
(0.0460)
-0.000369
(0.0529)
0.322***
(0.114)
0.133
(0.0985)
0.589***
(0.147)
0.480***
(0.148)
-0.153*
(0.0913)
-0.125
(0.0930)
0.0396***
2SLS Model (Dummy for Rent
in)
10.79***
(0.602)
0.112
(0.542)
0.113***
(0.0316)
0.0881
(0.291)
-0.0322
(0.0634)
-0.0359
(0.0513)
-0.0328
(0.0557)
0.237**
(0.120)
0.0732
(0.104)
0.685***
(0.146)
0.522***
(0.148)
-0.220**
(0.0938)
-0.155*
(0.0941)
0.0369***
2SLS Model (Value
of Rent in)
10.82***
(0.544)
0.0111
(0.0377)
0.113***
(0.0317)
0.114
(0.291)
-0.0329
(0.0635)
-0.0376
(0.0510)
-0.0344
(0.0556)
0.233*
(0.121)
0.0699
(0.104)
0.684***
(0.146)
0.522***
(0.148)
-0.219**
(0.0939)
-0.156*
(0.0943)
0.0368***
Variables
Log-Log (Dummy for
Rent in)
Log-Log (Value of
Rent-in)
2SLS Model (Dummy for Rent
in)
2SLS Model (Value
of Rent in)
(0.00975)
0.0326***
(0.00975)
0.0327***
(0.0103)
0.0430***
(0.0103)
0.0431***
(0.00994)
0.363***
(0.0954)
0.233***
(0.0641)
0.162
(0.119)
0.191***
(0.0651)
0.359*
(0.212)
-0.106
(0.243)
0.378*
(0.208)
0.00743
(0.00994)
0.362***
(0.0954)
0.233***
(0.0641)
0.162
(0.119)
0.191***
(0.0651)
0.359*
(0.212)
-0.107
(0.243)
0.375*
(0.208)
0.00752
(0.0100)
0.254***
(0.0969)
0.167**
(0.0681)
0.244**
(0.119)
0.181**
(0.0776)
0.115
(0.221)
0.234
(0.250)
0.260
(0.264)
(0.0100)
0.254***
(0.0969)
0.166**
(0.0684)
0.243**
(0.119)
0.184**
(0.0773)
0.115
(0.221)
0.232
(0.250)
0.248
(0.259)
-
-
(0.0437)
-0.00111
(0.0395)
-0.167***
(0.0422)
-0.0327
(0.0595)
(0.0437)
-0.00156
(0.0395)
-0.167***
(0.0422)
-0.0323
(0.0595)
-
-
-
-
-
-
Irrigations
Number of Ground
Irrigations
Plot Located at Head
Plot Located at Middle
Village Electrification
Internal Developed Road
Cotton Grower
Rice Grower
Sugarcane Grower
Distance Nearest Weekly
Market
Distance Nearest Bank
Distance Nearest City
Distance District Katcheri
Variables
Distance Tehsil Katcheri
District Dummies
Observations
R-squared
Log-Log (Dummy for
Rent in)
-0.0171
(0.0461)
Yes
Log-Log (Value of 2SLS Model (Dummy for Rent
Rent-in)
in)
-0.0163
(0.0462)
Yes
Yes
Test Statistics
1120
1120
0.514
0.478
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
2SLS Model (Value
of Rent in)
-
1120
0.514
1120
0.478
Yes
(Note: Refer to Table 5 in the Appendix to confirm that the residuals are normally distributed)
The table above indicates that most of the site characteristics have a correlation with perceived plot value per acre. The site
characteristics also have the expected sign; for example fertile land, no soil erosion, number of irrigations by canal and ground water,
plot located at head and middle of water course all have a positive correlation with the dependent variables across most of the model
versions. Similarly, waterlogging and salinity have a negative correlation on perceived value of land per acre although salinity is not
statistically significant across all of the models. Four of the coefficients for development variables are correlated with the dependent
variable. Access to electricity, internal road, cotton growers, and sugarcane growers are positively correlated with perceived land
values across most of the model versions. Variables for access to electricity and cotton growers were chosen to act as a proxy for other
variables which captured the effect of an increase in income or development of a village. Similarly dummies for cotton growers and
sugarcane growers were chosen to capture the effect of an increase in prices of crops and choice of crop. Most of the physical
variables were not correlated with the dependent variable with the exception of distance to nearest city. Demographic variables such as
the age of the respondent or if the respondent has ever attended school do not appear to have any correlation with the dependent
variable, however average mauza income does have a positive correlation with perceived land value in some of the model versions.
Model results indicate that ownership status and renting of plots are not correlated with land value per acre.
Another way of looking at the results is by classifying the results that have a correlation with perceived land value at the 1%
level. Three of these variables have an impact on soil fertility and erosion (quality of land) and four of these are related to water
quality and access. This again shows the importance of access and maintenance of good quality land and water as well as the
importance
of
physical
characteristics
such
as
waterlogging
and
salinity.
VII. Conclusion and Policy Implications
Land disputes and development of land markets has proven to be a difficult task in the
history of Pakistan. Land is an important social and financial asset and previous literature has
proven that an increase in productivity of agricultural land will lead to a decrease in poverty. Yet
factors which increase agriculture productivity also indirectly increase the value of land.
Therefore, this study attempts to fill a gap in previous research by identifying what factors affect
agricultural land values in rural areas of Pakistan. Finding out what factors affect perceived value
of land per acre in rural Pakistan could help the Government of Pakistan decide what to invest in
to promote development of rural land. This in turn will make rural land more attractive to local
and foreign investors which will increase investment and development and decrease poverty.
Our research study used a hedonic regression model and two-staged least square model to
determine what demographic, site, development, or physical characteristics have a correlation
with the perceived value per acre of agricultural land. Four different versions of the model were
used to analyze the impact of rent vs value of rent and to counter potential issues of endogeneity.
Data for this study was obtained from the Pakistan Strategy Support Program’s Rural Household
Panel Survey (RHPS) of 942 households across 19 districts and 3 provinces who are currently
involved in agriculture. Overall, the results are consistent with international literature on the
subject. Model results indicate that most of the site and physical characteristics are correlated
with perceived land value and only a few of the development indicators and none of the
demographic variables have a correlation with perceived land value. Specifically fertile land,
lack of soil erosion, number of canal and ground water irrigations, location of plot at head and
middle of watercourse, access to electricity, internal road, cotton grower, sugarcane grower, and
average mauza income are positively correlated with perceived land value per acre.
Waterlogging, salinity, and distance to nearest city are negatively correlated with perceived land
value.
These results provide some important policy implications which the Government of
Pakistan can consider. It is worth noting that the Government of Pakistan has already announced
a series of reforms to boost the economy and development. One of the main themes of the Vision
2025 of the Government and the Planning Commission include modernization of infrastructure
and regional initiatives. Results from the model above suggest that the improvement of
infrastructure and usage of regional initiatives could have the desired effect of increasing
agriculture productivity through land development. For example, the results on site and physical
characteristics suggest that the quality of land and how the farmer uses the land is important for
productivity. Most of the site and physical characteristics are correlated with perceived land
value which indicates that the Government of Pakistan should initiate policies which promote
regional or rural initiatives which improve water access and maintenance of watercourses.
Fertility level of land, soil erosion, and salinity can be controlled to a certain extent by sufficient
access to good quality water and proper maintenance of the land and watercourses. Therefore, if
the Government provides training and other resources at the micro-level, farmers will have a
better chance to ensure that their land is of high quality and water access is sufficient to ensure
higher productivity. Similarly access to electricity is important and the Government should
ensure access to electricity throughout Pakistan.
One of the limitations of the model used above is that we could only include
characteristics which were measurable or observable (ex. access to road, access to water, soil
quality etc.). Theoretically, there are other variables that could be correlated with perceived land
value. For example, implementing institutional rules which improve good governance, land
titling policies, inheritance policies, and promoting ownership of land by foreigners could all be
positively correlated with land values. All of these factors could be used as mechanisms to
achieve the desired objective to promote agriculture productivity and development in rural areas.
Lastly, considering the development needs of Pakistan, the area of land ownership and
development of rural agricultural land cannot be ignored. Similarly, proper investment into rural
areas can turn them into centers of commerce which will boost productivity and economic
growth. In the long run, this will improve investment and competition in the area.
VIII. References
Aftab, Safiya, Ali Raza Bhutta, Maham Farhat, Stephen Jones, and Mujib Khan “Political
Economy Analysis of the Land Record System” July 2012, Oxford Policy Management
publication
Bover, Olympia, and Pilar Velilla. “Hedonic house prices without characteristics: the case of
new multiunit housing”. No. 3140. Centre for Economic Policy Research, 2002.
Cavailhès, J. & Wavresky, P. 2003. Urban influences on peri-urban farmland prices. European
Review of Agricultural Economics 30:333-357.
Ciaian, Pavel, et al. Institutional Factors Affecting Agricultural Land Markets. No. 118. Centre
for European Policy Studies, 2012.
Deininger, Klaus. "Land policies for growth and poverty reduction: key issues and challenges
ahead." Inter-Regional Special Forum on the Building of Land Information Policies in the
Americas, Aguascalientes, Mexico. 2004.
Deininger, Klaus, Hilhorst, Thea, and Songwe, Vera. “ Identifying and Addressing Land
Governance Constraints to Support Intensification and Land Market Operations: Evidence from
10 African Countries” Food Policy Volume 48, October 2014 Pg 76 – 87
Du, Hongbo, and Corinne Mulley. "The short-term land value impacts of urban rail transit:
quantitative evidence from Sunderland, UK." Land Use Policy 24.1 (2007): 223-233.
Food and Agriculture Organization “Overview of Land Value Conditions” Rome 2003, United
Nations publication.
Gardner, B., and C. Nuckton. "Factors affecting agricultural land prices." California Agriculture
33.1 (1979): 4-6.
Government of Pakistan Publication “Agricultural Census of Pakistan 2010”, available at the
Federal Bureau of Statistics
Government of Pakistan Publication “Agricultural Statistics of Pakistan 2010-11” available at
the Federal Bureau of Statistics
Guiling, Pam, B. Wade Brorsen, and Damona Doye. "Effect of urban proximity on agricultural
land values." Land Economics 85.2 (2009): 252-264.
Hirashima, Shigemochi. "The land Market in Development: a case Study of Punjab in Pakistan
and India." Economic and Political Weekly (2008): 41-47.
Mumtaz, Khawar, and Meher M. Noshirwani. "Women’s Access and Rights to Land and
Property in Pakistan." International Development Research Centre. Available at: http://www.
shirkatgah. org/Women_access-rights-to_land_property_in_Pakistan. pdf (2006).
Naqvi, Syed Nawab Haider, Mahmood Hasan Khan, and Muhammad Ghaffar Chaudhry.
"Structural change in Pakistan's agriculture." Structural change in Pakistan's agriculture (1989).
Qureshi, Madeeha G., Sarfraz Khan Qureshi, and Abdul Salam. "Impact of Changing Profile of
Rural Land Market in Pakistan on Resource Allocation and Equity [with Comments]." The
Pakistan Development Review 43.4 (2004): 471-492.
Pakistan Economic Survey 2012-13: Available at http://finance.gov.pk/survey_1213.html.
Accessed October 28, 2013
PANOS South Asia Study “Leveling the Playing Field: A Survey of Pakistan’s Land Reforms”
March 2011 published as part of project by Swedish International Development Agency (SIDA)
Peterson, W. L. 1984. Land Quality and Prices. University of Minnesota. Institute of Agriculture,
Forestry and Home Economics. Staff Paper P84-29:1-34.
Peterson, W. L. 1986. Land Quality and Prices. American Journal of Agricultural Economics
68:812-819.
Plantinga, Andrew J., and Douglas J. Miller. "Agricultural land values and the value of rights to
future land development." Land Economics 77.1 (2001): 56-67.
Renkow, Mitch. "Land prices, land rents, and technological change: evidence from Pakistan."
World Development 21.5 (1993): 791-803.
Roberts, Michael J., Barrett Kirwan, and Jeffrey Hopkins. "The incidence of government
program payments on agricultural land rents: The challenges of identification." American
Journal of Agricultural Economics 85.3 (2003): 762-769.
Rosen, S. 1974. Hedonic Prices and Implicit Markets: Product Differentiation in Pure
Competition. Journal of Political Economy 82:34-55.
Saita, Yumi. Land Prices in the Tokyo Metropolitan Area: A Hedonic Analysis of Judicial
Auction Prices. No. 03-E. Bank of Japan Working Paper Series, 2003.
Stillman, Steven. "Examining changes in the value of rural land in New Zealand between 1989
and 2003." (2005).
Taylor, Mykel R., and Gary W. Brester. "Noncash income transfers and agricultural land values."
Applied Economic Perspectives and Policy 27.4 (2005): 526-541.
USAID Publication - USAID Country Profile – Property Rights and Resource Governance –
Pakistan 2011
Vasquez, O., Wright, K. S., Nelson, J. R. & Hamilton, J. R. 2002. Determining the effects of
land characteristics on farmland values in south-central Idaho. Paper presented at AAEA –
WAEA Annual Meeting in Long Beach, CA. University of Idaho. Department of Agricultural
Economics and Rural Sociology. Research Series No. 02-05:1-18.
Vural, Hasan, and Halil Fidan. "Land marketing and hedonic price model in Turkish markets:
Case study of Karacabey district of Bursa province." African Journal of Agricultural Research
4.2 (2009): 71-75.
IX.
Appendix
Table 3: Description of Sample from Round 1.5 of PSSP Rural Household Survey
Province
Punjab
Punjab
Punjab
Punjab
Punjab
Punjab
Punjab
Punjab
Punjab
Punjab
Punjab
Punjab
Sindh
Sindh
Sindh
Sindh
Sindh
KPK
KPK
Total
District
Attock
Bahawalnagar
Bhakkar
DG Khan
Faisalabad
Jhang
Kasur
Khanewal
Multan
Rahim Yar Khan
Sargodha
Vehari
Hyderabad
Jacobabad
Sanghar
Thatta
Dadu
Mansehra
Nowshera
Description of the Variables:
Log of Perceived Value of Land Per Acre
Number of Households
16
58
78
42
43
55
39
45
22
42
27
54
57
86
26
86
50
45
71
942
This variable is the dependent variable of the model and asks the farmer what is their perceived
value of the land if it was sold today. The total value of the land was divided by the number of
acres owned by the household. The natural log of this new variable was used.
Log of Respondent’s Age
This variable takes the log of the farmer’s age. The expected sign is positive
Average Mauza Income
This variable provides the annual average income per person in a given mauza
Dummy for if the Respondent has Ever Attended School
This dummy variable assigns a value of “1” if the farmer has ever attended school, otherwise the
value is “0.” The expected sign is positive.
Dummy for Ownership of Plot
This dummy variable assigns a value of “1” if the farmer owns the plot that they are managing.
Expected sign for this variable is positive.
Dummy for Renting a Plot
This dummy variable assigns a value of “1” if the farmer rent-in a plot that they are managing.
Expected sign of this coefficient is positive.
Value of Rent
This variable provides the annual value of rent per acre for a given plot
Dummy for Mechanization
This dummy variable assigns a value of “1” if the household used a laser land leveler, tractor, or
thresher on their plot. Otherwise the value was “0.” The expected sign is positive.
Dummy for Flat Land
If a farmer’s land is flat this variable has a value of “1” otherwise the value is “0.” Expected sign
of the coefficient is positive.
Dummy for Terraced Land
If a farmer’s land is terraced this variable has a value of “1” otherwise it has a value of “0.”
Expected sign of the coefficient should be negative.
Dummy for Fertile Land
If a farmer’s land is fertile this variable has a value of “1” otherwise it has a value of “0.”
Expected sign of the coefficient is positive.
Dummy for Moderate Fertile Land
If a farmer’s land is moderately fertile this variable has a value of “1” otherwise it has a value of
“0.” Expected sign of the coefficient is uncertain but probably negative.
Dummy for No Soil Erosion
If there is no soil erosion on farmer’s plot, this variable is assigned a value of “1” otherwise the
value is “0.” Expected sign of the coefficient is positive.
Dummy for Mild Soil Erosion
If there is some soil erosion on farmer’s plot, this variable is assigned a value of “1” otherwise
the value is “0.” Expected sign of the coefficient is unclear but probably negative.
Dummy for Salinity
If the soil on a plot is saline, this variable is assigned a value of “1” otherwise the value is “0.”
Expected sign of the coefficient is negative.
Dummy for Waterlogging
If the soil on a plot is waterlogged, this variable is assigned a value of “1” otherwise the value is
“0.” Expected sign of the coefficient is negative.
Dummy for Plot at Head
If a farmer’s agricultural plot is at the head of the watercourse, this variable is assigned a value
of “1” otherwise the value is “0.” Expected sign of the coefficient is positive.
Dummy for Plot at Middle
If a farmer’s agricultural plot is in the middle of the watercourse, this variable is assigned a value
of “1” otherwise the value is “0.” Expected sign of the coefficient is positive.
Number of Canal Water Irrigations
Log of the number of times the farmer used canal water to irrigate the plot. Expected sign should
be positive.
Number of Ground Water Irrigations
Log of the number of times the farmer used ground water to irrigate the plot Expected sign
should be positive.
Dummy for Non-Farm Income
If a farmer is earning income from a source other than agriculture, this variable is assigned a
value of “1” otherwise the value is “0.” Expected sign should be positive.
Dummy for Village Electrification
If electricity is available in the mouza, this variable is assigned a value of “1” otherwise the value
is “0.” Expected sign of the coefficient is positive.
Dummy for Internal Road
This variable has a value of “1” if the village has an internal road made of gravel, asphalt, bricks,
or concrete. For other materials used for the road (example mud) and if there is no internal road
this variable has a value of “0.” Expected sign of the coefficient is positive.
Dummy for Gas Cylinder
If gas cylinder is available in the mouza, this variable is assigned a value of “1” otherwise the
value is “0.” Expected sign of the coefficient is positive.
Dummy for Cotton Grower
This variable has a value of “1” if the farmer grows cotton on the plot that they manage.
Expected sign of the coefficient is positive.
Dummy for Rice Grower
This variable has a value of “1” if the farmer grows rice on the plot that they manage. Expected
sign of the coefficient is positive.
Dummy for Sugarcane Grower
This variable has a value of “1” if the farmer grows sugarcane on the plot that they manage.
Expected sign of the coefficient is positive.
Log of Nearest City
This variable takes the log of the distance from the village to the nearest city in kilometers.
Expected sign of the coefficient should be negative.
Log of Nearest Output Market
This variable takes a log of the distance from the village to the nearest Output market. Expected
sign of the coefficient is negative.
Log of Bank
This variable takes a log of the distance from the village to the nearest bank. Expected sign of the
coefficient is negative.
Log of District Mandi
This variable takes a log of the distance from the village to the nearest district market. Expected
sign of the coefficient is negative.
Log of Tehsil Katcheri
This variable takes a log of the distance from the village to the nearest tehsil headquarters.
Expected sign of the coefficient is negative.
District Dummys
This variable is a dummy for each of the given districts in the sample
.6
.4
Density
.6
0
0
.2
.2
.4
Density
.8
1
.8
Table 4: Distribution of Residuals
-4
-2
0
2
-4
-2
Residuals
2
0
2
.8
1
.4
Density
.6
.8
.6
.2
.4
0
.2
0
Density
0
Residuals
-4
-2
0
Residuals
2
-4
-2
Residuals